Re: [Libvir] PATCH: Make the virBuffer API harder to misuse

On Mon, Apr 28, 2008 at 02:59:06PM +0100, Daniel P. Berrange wrote: > On Mon, Apr 28, 2008 at 08:19:20AM -0400, Daniel Veillard wrote: > > On Mon, Apr 28, 2008 at 01:13:00PM +0100, Daniel P. Berrange wrote: > > > On Mon, Apr 28, 2008 at 03:17:33AM -0400, Daniel Veillard wrote: > > > > I would prefer to relax the 'non-public' point and let the compiler > > > > compute the size in some ways rather than hardcode based on a word size > > > > indication which may not take into account specific alignment problems > > > > on some platforms. > > > > > > One other option I considered is to just define the struct in the public > > > header with meaningless field names > > > > > > struct _virBuffer { > > > char a; > > > > char *a; > > > > > unsigned int b; > > > unsinged int c; > > > }; > > > > > > The real version is re-declared with proper names in buf.c, so this will > > > at least discourage its use which is probably good enough. > > > > Fine by me. > > Here's the updated patch... Okay, that's massive, best to apply it early since we agree on it.

> + > +/* If adding more fields, ensure to edit buf.h to match > + the number of fields */ > +struct _virBuffer { > + unsigned int size; > + unsigned int use; > + unsigned int error; > + char *content; > +}; I would still prefer if we could get the compiler to check the structure (or at least their size) is the same... but that's should not impact how the client code will be using the buffers so, we can try to fix that after the main commit.